Re: Kernel BUG on mounting BtrFS / after reboot

[Chris Mason <chris.mason@oracle.com>]

On Thu, Feb 18, 2010 at 05:38:22PM +0000, Alex Elsayed wrote:
> Chris Mason <chris.mason <at> oracle.com> writes:
> 
> > 
> > Ugh ok.  The first thing I'd like to do is give you a patch to
> > completely disable the tree log replay and give you the chance to backup
> > critical data.
> > 
> > Do you already have a backup of the critical things on this drive?  The
> > problem you're hitting is that tree-logging code is trying to replay an
> > fsync of a file, but it can't find the file to replay.  This could be a
> > small and localized corruption or it could be a larger problem.  We'll
> > have to fix things one at a time to figure it out.
> > 
> > The easiest way to move forward would be to save a complete copy of the
> > FS with dd, but that's probably not very easy given the size.
> > 
> > -chris
> 
> Well, I ran badblocks on the drive, and it was happily silent, so that's a good 
> sign. I alo attached btrfsck output upthread, which says that there are 44 
> inodes with errors. Unfortunately, I don't really have a drive big enough to 
> back up to. I may be able to borrow one from a friend, though.

I think the btrfsck output is missing.  It sounds like we'll survive if
we just skip this part of the log replay.  I'll cook a patch based on
the btrfsck output.

-chris